Predicting Peak Prices in the Current Day-Ahead Market

Abstract

The work presented has studied price developments in the day-ahead market. The statistical properties inherent in the time series constituting the day-to-day prices have been investigated. It is shown that these properties are radically different in 2022 than former years and resemble more chance like properties for which make predicting future peak prices especially hard. To overcome this, a space domain approach was adopted to determine whether information about co-variant price zones could improve predictions. By using a multi-variate regression method, it is possible to accurately predict prices in NO2 using information about concurrent prices in other price zones. However, an attempt to use historic prices in multiple price zones to predict future prices gave only modest results. One reason for this is the high degree of entropy underlying price developments in the day-ahead market in 2022.